[QUOTE=LaurV;440342]Reserving it. Ran a srbsieve on it and there are only about 1000k remaining at n=2k, in less than one hour, so I will take it to 25k or how much it will reach over the weekend. No need save files from Christian, and thanks him that he came back to unreserve it. Other guys would just go MIA and abandon, and we would never knew...[/QUOTE]

This base is in the recommended bases thread and had already been searched to n=2500. A list of k's remaining at n=2500 is attached to the first post in that thread. But if you are already at n=2000 then you may as well continue with what you are doing.

I predict you will not get to n=25K over the weekend. lol I'll reserve it to n=10K for you for the time being.

[QUOTE=gd_barnes;440343]I predict you will not get to n=25K over the weekend. lol I'll reserve it to n=10K for you for the time being.[/QUOTE]
Your prediction was right, the weekend is almost over and I am close to n=6k only. My "estimation" of the speed was off by an order of magnitude, because I forgot the fact that, comparing with R66 on which my experience is based, this one grows 10 times faster with every N, so therefore testing for primality becomes much slower much earlier. I will however continue to at least n=10k. This was a single core job up to now. I am going to split it in 2 or maybe 3 cores.

I have a problem here... Huston? Help!

What is with 3364*672^n-1?

This 3364 is 2^2*29^2, and all the even powers are algebraically factorable (as x^2-1) and are correctly eliminated by the srsieve (not by newpgen, however, and that was where my investigation started, I wondered why the difference).

But on the other hand, all odd powers should be divisible with 673, because if I add and subtract 3364, I get 3364*672+3364-3364-1=3364*673-3365=3364*673-5*673, etc.

Did I just jumped ahead in proving R672 by finding a smaller Riesel number? :razz:

[QUOTE=LaurV;440430]I have a problem here... Huston? Help!

What is with 3364*672^n-1?

This 3364 is 2^2*29^2, and all the even powers are algebraically factorable (as x^2-1) and are correctly eliminated by the srsieve (not by newpgen, however, and that was where my investigation started, I wondered why the difference).

But on the other hand, all odd powers should be divisible with 673, because if I add and subtract 3364, I get 3364*672+3364-3364-1=3364*673-3365=3364*673-5*673, etc.

Did I just jumped ahead in proving R672 by finding a smaller Riesel number? :razz:[/QUOTE]
Look at exclusion 2 on the following page:
[URL]http://www.noprimeleftbehind.net/crus/Riesel-conjectures.htm[/URL]

Base 12 has a similar issue.

Ok, this brings a lot of light! Thanks.
I guess that the right action for me now is to [U]manually[/U] remove the tricky k from the list, and don't waste time with sieving and LLR-ing/pfgw-ing it. Actually only sieving it, because it never survives the sieving.

Lots of bases have k's that have algebraic factorization to remove some n's and where a trivial factor removes all remaining n's to effectly remove the k from the conjecture. I do my best to pick those out on the pages when a base is reserved. You'll notice on the main Riesel page that I show the statement: "k = 3364 proven composite by partial algebraic factors." Fortunately there are relatively common patterns that we have come up with to determine such k ahead of time on most bases.

You are right. The correct action is to manually remove the k from your search.

S832 tested to n=200k (100-200k)

nothing found, 2 remain

Results emailed - Base released

Reserving R783 to n=200k (100-200k) for BOINC

Reserving S625 to n=100k (25-100k) for BOINC

You can save some CPU time if you reformat the sieve file before sending it to compute nodes.
Compare (note: this is the same job):
[CODE]/home/serge/NumTheory/S625> llr -d t1
Base prime factor(s) taken : 5
Starting N-1 prime test of 3068*5^100021+1
Using all-complex FMA3 [COLOR="SeaGreen"]FFT length 18K[/COLOR], Pass1=384, Pass2=48, a = 3
3068*5^100021+1 is not prime. RES64: 24E72669D12B7D16. OLD64: 6EB5733D7382773F
[COLOR="SeaGreen"]Time : 14.867 sec.[/COLOR]
/home/serge/NumTheory/S625> llr -d t2
Base prime factor(s) taken : 5
Starting N-1 prime test of 15340*625^25005+1
Using zero-padded FMA3 [COLOR="Red"]FFT length 35K[/COLOR], Pass1=448, Pass2=80, a = 3
15340*625^25005+1 is not prime. RES64: 24E72669D12B7D16. OLD64: 6EB5733D7382773F
[COLOR="Red"]Time : 28.478 sec.[/COLOR]
[/CODE]

Here is the script:
[CODE]echo '15000000000000:P:1:5:257' > sieve-S625-25K-100K.txt
awk 'NF>1{k=$1;n=$2*4;while(k%5==0){k/=5;n++} print k, n}' sieve-sierp-base625-25K-100K.txt >> sieve-S625-25K-100K.txt
[/CODE]
Looks simple, right? ...Works wonders.

[QUOTE=Batalov;441219]You can save some CPU time if you reformat the sieve file before sending it to compute nodes.
[/QUOTE]

I dont understand the script. What do you mean with reformat?